So I've had intermittent issues over a long time of one of my four nodes being fenced and going unresponsive. I think I've narrowed down the issues to some basic network issues that have never really dealt with and I'm trying to figure out best practices.
I am told that "storage communication should never be on the same network as corosync!" - generally speaking, this is exactly what I've done with all of my nodes - lots of traffic, all on the same 172.16.120.0 network.
I've tried to rectify this issue, but I'm not sure if what I've done will be sufficient.
My cluster nodes all have two nics - 1-1G and 1-2.5G. Initially, all traffic has been put through the 2.5G nics. To try and separate out the Corosync network, I moved the management interface from the 2.5G nic to a new vlan connected bridge port bridge on the 1G nic. (I followed this post).
Now, I believe that all of my corosync traffic will be moving across that second 1G nic while VMs and data traffic remain on the 2.5G nic.
Will this setup achieve that goal of not having storage communication on the same network as corosync? The VMs and management interface are still on the 172.16.120.0 network, but management is on a completely different nic.
If it helps, here is my /etc/pve/corosync.conf:
And here is /etc/network/interfaces:
And finally, here is what a node generally looks like:
I am told that "storage communication should never be on the same network as corosync!" - generally speaking, this is exactly what I've done with all of my nodes - lots of traffic, all on the same 172.16.120.0 network.
I've tried to rectify this issue, but I'm not sure if what I've done will be sufficient.
My cluster nodes all have two nics - 1-1G and 1-2.5G. Initially, all traffic has been put through the 2.5G nics. To try and separate out the Corosync network, I moved the management interface from the 2.5G nic to a new vlan connected bridge port bridge on the 1G nic. (I followed this post).
Now, I believe that all of my corosync traffic will be moving across that second 1G nic while VMs and data traffic remain on the 2.5G nic.
Will this setup achieve that goal of not having storage communication on the same network as corosync? The VMs and management interface are still on the 172.16.120.0 network, but management is on a completely different nic.
If it helps, here is my /etc/pve/corosync.conf:
Code:
ogging {
debug: off
to_syslog: yes
}
nodelist {
node {
name: prox4
nodeid: 4
quorum_votes: 1
ring0_addr: 172.16.120.36
}
node {
name: prox5
nodeid: 3
quorum_votes: 1
ring0_addr: 172.16.120.37
}
node {
name: prox7
nodeid: 2
quorum_votes: 1
ring0_addr: 172.16.120.34
}
node {
name: pve1z440
nodeid: 1
quorum_votes: 1
ring0_addr: 172.16.120.39
}
}
quorum {
provider: corosync_votequorum
}
totem {
cluster_name: congregation
config_version: 12
interface {
linknumber: 0
}
ip_version: ipv4-6
link_mode: passive
secauth: on
version: 2
}And here is /etc/network/interfaces:
Code:
auto lo
iface lo inet loopback
iface enp1s0 inet manual
iface eno1 inet manual
auto vmbr0
iface vmbr0 inet manual
bridge-ports enp1s0
bridge-stp off
bridge-fd 0
bridge-vlan-aware yes
bridge-vids 2-4094
auto vmbr2
iface vmbr2 inet manual
bridge-ports eno1
bridge-stp off
bridge-fd 0
bridge-vlan-aware yes
bridge-vids 2-4094
auto vmbr2v120
iface vmbr2v120 inet static
address 172.16.120.36/24
gateway 172.16.120.1
bridge-ports eno1.120
bridge-stp off
bridge-fd 0
source /etc/network/interfaces.d/*And finally, here is what a node generally looks like:
